How Kubernetes plays Tetris with your containers - Salman Iqbal - NDC Oslo 2021

Kubernetes does an excellent job of abstracting away the infrastructure and efficiently deploying containers to improve resource management. The Kubernetes scheduler is a core component of the control-plane. It is in charge of deciding where the Pods are placed in your cluster. You can think about the Kubernetes scheduler as a skilled Tetris player that always tries to optimise the placement of your containers. But how does it decide what is the *best* node? Does it prioritise big Nodes or smaller ones? Can you help the scheduler decide? In this talk we will cover: - The internals of the Kubernetes scheduler and discover how Nodes are filtered before pods are assigned to them - Tools at your disposal if you want to influence the scheduler to place workloads on specific nodes such as GPU, deploy applications to reduce latency and more!